IRC channel logs
2023-06-17.log
back to list of logs
<stikonas>they seem to say their situation is better though <stikonas>oriansj: I've just updated stage0-posix submodules to pull in those mescc-tools fixes that you merged recently <stikonas>oriansj: do we have any other blockers for the release? <stikonas>(I think janneke is basically waiting for a new stage0-posix release before releasing mes 0.25) <oriansj>stikonas: just me overthinking a single change in M1 to properly do range checks for immediates <stikonas>ok, that sounds like it might be worth including <kerravon>so how do you actually make a computer that has switches, given that they are no longer sold new? <kerravon>and is that computer considered to be secure for bootstrapping purposes? <stikonas>kerravon: but if you think, no modern computer can be truly bootstrapped unless it has switches <stikonas>cause how are you going to put anything on it <oriansj>kerravon: making a computer only takes a handful of chips <stikonas>possibly you can electrically talk to SPI too <stikonas>or to be more precise, communication protocol with that media <oriansj>such as Compact Flash but SPI is a protocol for using very few wires <oriansj>to get multiple different chips to talk to each other <kerravon>switches make you lose your data if you want to do it again. punched cards stay <stikonas>that's correct. Anyway, when I was saying "switches", I meant more generally those primitive input methods <oriansj>kerravon: yes and they also have the property that nothing gets between you can what you want in RAM or EPROM <stikonas>(but not something like booting computer from USB) <stikonas>and that is right now a harder problem that we have solved <stikonas>at the moment we can bootstrap on POSIX or on BIOS, so we assume at least System Firmware <stikonas>(there is also some incomplete work done on UEFI bootstrap which is somewhere between BIOS and POSIX bootstrap) <oriansj>but I am going to chase the bootstrap chain down to building our own custom firmware and hardware <oriansj>after that we run into a technology limit; unless someone is really good at chemistry and wants to help get a RYF lithography printer we can use to make our own chips safely at home <oriansj>ekaitz: well it is a *VERY* clean design and certainly possible to be productive on it, it just has some things that make one really appreciate even the worst POSIX kernel available <ekaitz>Oberon is great! I love the way Wirth justifies things <ekaitz>the introduction to the book is something like: I was coding and I realized the computer was too complex <ekaitz>things I like from oberon: almost no multitasking, single user approach <ekaitz>that should be the default on some devices <ekaitz>users and permissions are too much for many things <kerravon>stikonas - what is this harder problem that has been solved and how? <kerravon>how come this SPI doesn't have a ground? <kerravon>is a punched card reader the most appropriate technology to be using so that you can see the data? are they sold new anymore? or can they be made? <kerravon>another possibility is to burn a commercial CDROM when you believe everything is in order, and then use a microscope to read the pits <muurkha>maybe they found a wiki full of illegal information <muurkha>kerravon: it's pretty straightforward to wire switches up to a CPU and RAM. Ben Eater has some YouTube videos doing this with a 6502, and someone [maybe Ben Eater] also has something similar using an AVR8 <pabs3>from another channel: <someone> hearing rumors that their sysadmin is resigning in the next 24 hours and will not help with migration efforts <someone> *miraheze's <muurkha>SPI is a real pain to do by hand but it is feasible <muurkha>oh, I see stikonas answered with stuff about Ben Eater's divulgation work <muurkha>oriansj: what are the things about Oberon that make one appreciate POSIX? <muurkha>kerravon: SPI does have a ground, it's single-ended, not differential <muurkha>sometimes people don't draw the ground in a circuit diagram but it's there <muurkha>dammit. is there a convenient way to get wget to ignore the nofollow attribute? <muurkha>and I guess --reject-regex will tell wget to not try to mirror all the edit§ion=13 and Special:UserLogin&returnto= pages <stikonas>kerravon: I just meant that it's quite a bit harder to bootstrap without any firmware (though still much easier than building/bootstrapping hardware) <oriansj>muurkha: well for one the editor is bad relative to vi/mg; there is no C nor lisp available. The supported assembly language is just plain bad. <muurkha>oriansj: hmm, does it not have search and replace, or is the complaint that you have to take your hands off the home row? <oriansj>Not even a FORTH has been written for it (that should tell you something serious) <muurkha>FORTH isn't really compatible with the Oberon approach <muurkha>it might be useful to make a physically much smaller version of punched paper tape, maybe with a Hamming code and the possibility of sensing it optically instead of capacitively <muurkha>I mean optically or capacitively instead of conductively <oriansj>FORTHs are generally found everywhere, an operating system as old as Oberon should have a half dozen by now but it doesn't which tells you a great deal about how bad its assembly is <muurkha>I am not optimistic that my backup of the Miraheze Wiki will be successful, but I might make it <oriansj>muurkha: you don't need to backup the bootstrappable wiki <muurkha>oriansj: but Oberon relies on the compiler to keep different modules from crashing each other <muurkha>oh, do you have a full backup already? <kerravon>oriansj punched cards used in IBM mainframes (not sure about other systems) don't look like that airline thing <muurkha>it seems to be in Markdown rather than WikiText and omit the history <muurkha>kerravon: if you want to see how punched paper tape is encoded, install the bsdgames package and echo hello world | /usr/games/ppt <oriansj>muurkha: well it is the data that actually counts and if we as a community can stick to doing the wiki in markdown going forward we never have to worry about losing our wiki again <muurkha>oriansj: I think the Wiki also has some images in it that aren't in this repo <muurkha>Markdown is I think not a reasonable alternative to Wikimarkup but Commonmark might be <muurkha>GitLab has reasonable support for a Wiki using GLFM; wouldn't surprise me if sr.ht does too, but I don't know about it <muurkha>the crucial thing about Wiki is not so much which particular format it uses as the interaction model <oriansj>muurkha: well an append-only git repo which I will grant commit access to anyone who wants it with the assumption when someone cares more about it, that they will ultimately take it over and put it where they want <muurkha>this reminds me of when I tried to invent Twitter in 01997 by offering to create a mailing list for anyone who asked <oriansj>and as long as it remains a publicly clonable git repo (or other version control tool which has a free as in freedom implementation) I would be happy <muurkha>the crucial thing about a Wiki is that if you see a mistake you can click "Edit", click on the error, type your correction, and click "submit" <muurkha>you know for a long time WikiWikiWeb only had a single history item? <muurkha>if someone vandalized a page there was the option to revert it to the previous version <muurkha>but that was the entirety of its history and reversion functionality <muurkha>for like the first four or five years I think <muurkha>I think it's valuable to preserve the full history but it's not extremely important for bringing about the Wiki Way <muurkha>what is important is that the barrier to becoming a contributor is extremely low <pabs3>muurkha: there are better ways than just web mirroring btw. also, the ArchiveTeam folks are planning on putting crawling all miraheze wikis into web.archive.org <pabs3>oriansj: your copy seems to only have a few revisions? <pabs3>I tried to download the miraheze copy using git-remote-mediawiki, but hit a bug :( <kerravon>oriansj - actually this ties into something else i want to do. i want to preserve the entire source code to PDOS, plus the new bootstrapping code to get there, preferably for centuries. I was thinking of plastic punched cards - which I assume don't even exist. I believe plastic doesn't biodegrade so has the most chance of surviving. <oriansj>tfgbvhrdjeugdvbjbnectinudlbftienuuufrbffvjdt <oriansj>tfgbvhrdjeuggnvikjdnnhetbejcbrchkurifebiirnv <muurkha>oriansj: I think also the Talk pages are missing <kerravon>i have a program to count in binary. all you need to do is hit enter. but i haven't been able to get my 19 month old daughter to do that yet <muurkha>some plastics don't biodegrade some of the time <pabs3>I think the wikiteam tool is best, it can download a full dump of the wiki, and then you can use that to restore it elsewhere <muurkha>acetate in particular degrades enormously faster than paper, even bad paper <muurkha>some other plastics are much stabler than paper <muurkha>it seems like a lot of the Miraheze pages are already broken, which might be okay <oriansj>kerravon: the best method of preservation remains a collective group who dedicates effort to ensure the survival of the thing needing to be preserved; the bigger the number of people the better the odds <oriansj>muurkha: did anyone actually use the talk pages? <muurkha>not many that I've crawled so far, but a few <muurkha>collective groups are enormously less stable than baked clay tablets <muurkha>the collective group who dedicated effort to ensure the survival of the Ancient Egyptian religion were wiped out about 1600 years ago <oriansj>muurkha: true but are less likely to be broken into dust and used to make bricks <muurkha>likewise the collective group who dedicated effort to ensure the survival of the art of reading khipu (most of which were burned, but a few of which survive), the collective group who studied the teachings of Mozi, etc. <pabs3>on #faif, folks were talking about making a license that required you to distribute the software on clay tablets, for preservation :) <oriansj>but yes religions do have a bad problem with killing all competitors and wiping all trace of them off the planet <muurkha>and there are innumerable collective groups who have been forgotten so completely that we don't even know their names <muurkha>because they were illiterate, in most cases, though there are undeciphered scripts like Linear A <oriansj>which lack a rosetta stone to a language that survived <muurkha>but think of the Harappan civilization and the Kurgan culture. What stories did they tell each other? What plants did they know? <muurkha>nobody knows today, despite the collective groups who presumably worked hard to preserve that knowledge for centuries or millennia <kerravon>and maybe someone will send a package to the moon or to mars <kerravon>i don't trust anything i can't see though <kerravon>with a CD you can see the pits with a microscope <oriansj>well solar radiation will wreak any non-powered digital media storage <kerravon>but it's unclear to me whether it is viable <oriansj>but meteorites and lava will wreak anything <kerravon>solar radiation will wreck a pressed CDROM? <oriansj>kerravon: yes on a long enough time scale <kerravon>well, say 1000 years? if civilization hasn't recovered from a disaster by then, so be it <muurkha>yes, but a clay tablet needs a great deal less luck to survive than a collective group who dedicate effort to ensure survival <kerravon>do we not have some "modern clay" technology? <muurkha>clay and plastic each have advantages <muurkha>but the most important thing about plastic is *which plastic* <kerravon>at the same time as getting a commercially-pressed CDROM <muurkha>celluloid and acetate probably won't last a century, polyethylene and polypropylene and PET will probably last millennia <muurkha>commercially-pressed CD-ROMs are probably mostly polycarbonate, which is in between <kerravon>and can you create punch cards out of plastic? <oriansj>kerravon: you can do everything completely right but if some dick 20 years later decides to nuke something or pour acid on it, there isn't a whole bunch you can do to prevent that loss of knowledge <muurkha>oriansj: lots of copies keep stuff safeβ’ <kerravon>i would have multiple copies of my punched cards <muurkha>a thing to keep in mind about commercially-pressed CDs <kerravon>sure - it's not perfect. i just want to do something semi-maximable <muurkha>a vendor might fulfill your order by burning CD-Rs <oriansj>muurkha: yeah, which is why some people used gold tablets; only for other people to just melt them for the metal they really wanted <kerravon>can i verify that by using a microscope? <kerravon>ok, so not a problem. i'll warn the vendor in advance <oriansj>and grey goo might show up and render any strategy moot as the universe gets converted <kerravon>if the whole universe is destroyed, sure <kerravon>but there is the moon, mars, and potentially transmission to alpha centauri <muurkha>the reason religions wipe all trace of competitors off the planet is that the ones that do that have a much better chance of survival <oriansj>all backup plans ultimately are plans of what series of things must occur before the data will be lost <muurkha>so we tend to observe them more easily <kerravon>muurkha can i ask a commercial presser of cdroms to use the best possible plastic? <kerravon>the one you said is expected to last millenia <muurkha>you can't make CDs out of polyethylene or polypropylene <muurkha>PET is transparent but I don't think anyone has ever made a CD out of it <oriansj>the longest lasting discs currently are M-DISCs <muurkha>maybe there's a theoretical reason I don't know about, or maybe it's just a question of building a different CD molding machine <kerravon>but regardless - i want to see the pits with a microscope <oriansj>it is a data storage solution designed to last 1000 years <muurkha>I think the idea of making data microscopic is a pretty good one <muurkha>it reduces the materials costs involved by many orders of magnitude <muurkha>I mean maybe a dot-matrix printer has 9 pins and prints 6 lines per "inch", an obsolete term for 25.4 mm <ekaitz>oriansj: one of my students have a friend that uses those m-disks lol <muurkha>that gives you a dot pitch of 0.47 mm <muurkha>if you can reduce your dot pitch to 0.01 mm, 10 microns, then even on paper you can print 2200 times as many letters on the same paper readably <oriansj>ekaitz: yeah it is pretty popular with datahoarders who haven't gotten to the buy a tape drive for backups scale yet. <ekaitz>oriansj: this guy is a random dude that got anxiety when he realized he would lose his family pictures <kerravon>that mdisc link has under alternatives an "optical" which appears to be infinite <muurkha>if you can also make the paper thinner, like 10-micron-thick kitchen aluminum foil, that's another factor of 10. 22000 times denser <muurkha>a standard 600 dpi laser printer is 0.04 mm <oriansj>kerravon: well Syylex Glass Master Disc can in theory last forever in archival storage but cost 1000 Euro per disc <kerravon>i've been working on PDOS for nearly 30 years <kerravon>the man-hours is worth a hell of a lot more than 1000 euro * 7 continents or whatever <muurkha>I'm sorry I didn't know about it, that's a lot of work to go unnoticed <oriansj>kerravon: in which case I guess I should make a mirror <muurkha>or do you mean you're not sure why it hasn't gone viral? <oriansj>kerravon: do you have a master repo you prefer mirrors to be made from? <kerravon>yeah - more that it isn't viral as some sort of unrestricted at least teaching aid <muurkha>sometimes those things are just random <kerravon>yeah - regardless, i don't care that much. but i do want to preserve it <kerravon>it took something like 15 years to reconcile the mainframe <kerravon>so that i could have a mainframe version of it (z/pdos) <kerravon>15 years or more - depending on how you count <muurkha>right, I knew you meant the IBM 360 family <muurkha>but what do you mean about reconciling the mainframe? <muurkha>the WP page says that polycarbonate probably lasts a millennium <oriansj>wow, it has been a while since I've seen a fidonet address in a copyright header <muurkha>there's actually still an active fidonet network <oriansj>muurkha: oh indeed there is but much rarer these days <oriansj>the mainframe work looks quite interesting; <oriansj>I honestly never expected to see a publicly available MVS implementation <muurkha>I don't think it's an implementation of MVS, is it? <oriansj>muurkha: well if you look at the readme.txt; it appears it includes that <muurkha>I don't know enough about MVS to know what to make of its restrictions <oriansj>kerravon: it is now getting mirrored to 4 continents and I have submitted it to the software heritage foundation so it'll get into long term storage in another 2 continents <muurkha>does PDOS rely on the video card staying in text mode all the time? <muurkha>there's a note on the home page that suggests this <kerravon>and i have noticed that even in graphics mode, text still appears <kerravon>but why would you be in graphics mode without returning? <kerravon>i do go into graphics mode myself to display pictures <kerravon>someone added it, but i #defined it out due to bugs <muurkha>hmm, that might be part of why it's little-known <muurkha>I find it pretty valuable to be able to mirror a web site while I'm demonstrating my stupidity on IRC <muurkha>or compiling my code while I'm still editing it <muurkha>speaking of which I've downloaded 70 megabytes from bootstrapping.miraheze.org so far <muurkha>though it's pretty compressible, gzipping to only 13 <muurkha>and I have to admit I have my doubts about the value of backing up pages like bootstrapping.miraheze.org/wiki/Special:CentralAuth/πΊγ <muurkha>not to mention bootstrapping.miraheze.org/wiki/Special:CentralAuth/πππππππππππππππππππππππππππππππππππππππππππππ <kerravon>muurkha - sure, but it has taken 30 years just to get single-tasking to work. and i haven't even completed that to my satisfaction yet. we don't even have a public domain c90 compiler yet. the person who added multitasking support isn't around anymore to fix the bugs. plus that multitasking support took away things that i actually care about, like <kerravon>regarding MVS - depending on your definition, yes, z/pdos conforms to a subset of the MVS API <kerravon>oriansj - we have had MVS 3.8J as a publicly available MVS forever (ie 1983 or something) - doesn't that count too? <oriansj>kerravon: honestly the only MVS I have ever seen involved IBM contract terms <kerravon>after MVS 3.8J, all versions of MVS were copyrighted <oriansj>kerravon: but did any of the copies of the source code survive without an added copyright? <kerravon>oriansj - yes, sort of. The source control at the time wasn't what we would like, so the source and executables don't match exactly. Some people have tried to rectify that, including me, but in the end I decided it was easier to start from scratch with PDOS for the mainframe <kerravon>written in C instead of assembler and PL/S (which we don't even have the compiler for) <oriansj>kerravon: well PL/S was doomed to die simply because one couldn't even buy or rent a copy of a PL/S compiler and IBM legal went hard after RL/S for being "compatible" <kerravon>oriansj - i don't think PL/S was doomed to die. it was and is an internal IBM programming language, and they won't release it because they believe it gives them a competitive edge <oriansj>kerravon: I think that impression came from an internal engineer that was working with some smart customers and figured out how to get them to do most of their work for them and the customers were just happy that their fixes were incorporated promptly <oriansj>as it makes for a perfect excuse to keep sharing the source code with the customer and still claiming to have magical competitive advantages <oriansj>looking at the language features themselves; it doesn't appear to be that hard to implement <oriansj>(which explains why Rand literally gave away their implementation for free) <oriansj>But reading the code written in PL/S I see why IBM could have believed it gave them an advantage but they exposed too much of the compiler internals to direct manipulation. <kerravon>i don't know much about languages. i just use C and it seems fine to me. i doubt that i would be more productive using pl/s <oriansj>I'd skip learning it; C with inline assembly and linking assembly gives so much more flexibility and requires you do so much less book keeping in your code <muurkha>kerravon: yes, multitasking can be done without sacrificing features, but it's much more difficult <stikonas>though I'm still confused how kernel command line is populated when kexec-linux is called <Mikaku>stikonas: the kexec implementation in Fiwix includes a kernel parameter called 'kexec_cmdline=' where you set the command line to be passed to the new kernel <stikonas>so we basically set linux command line inside fiwix command line <Mikaku>you must instruct Fiwix that it will have to kexec another kernel <stikonas>that's fine. I think I remember seeing it before <muurkha>Mikaku: I don't know if I've said this often enough but Fiwix is amazing <stikonas>without kiwix we would be nowhere close to full kernel bootstrap that we have on x86 <Mikaku>muurkha, stikonas: that's really kind, thank you very guys <Mikaku>rickmasters is helping a lot to improve it